Notice: We have migrated to GitLab launching 2024-05-01 see here: https://gitlab.rtems.org/

Changes between Version 44 and Version 45 of GSoC/2013/ParavirtualizationOfRTEMS


Ignore:
Timestamp:
09/18/13 18:35:18 (11 years ago)
Author:
Phipse
Comment:

L4Re as a hypervisor to RTEMS

Legend:

Unmodified
Added
Removed
Modified
  • GSoC/2013/ParavirtualizationOfRTEMS

    v44 v45  
    440440For instance a time_warp function for the clock interrupt is needed.
    441441Claudio da Silva has provided one [https://gist.github.com/cdcs/5874932 here].
     442
     443=  More Hypervisors (L4Re) =
     444
     445
     446The concept of the virtualization layer should be portable.
     447I write should be, because I haven't tested it.
     448From my point of view, it should be an easy task to port the virtualpok BSP to L4Re.
     449L4Re provides a full vCPU interface and library functions to control it.
     450Compared to POK this is heaven.
     451
     452In my previous attempt to virtualize RTEMS on L4Re, before this project, I made L4Re load the RTEMS binary and then configure the vCPU to start at the binaries entry point, etc. .
     453This approach required the RTEMS binary to be analysed at compile-time of the L4Re application.
     454Also RTEMS was modified. Back then I introduced a new CPU target called l4vcpu, which was mostly a copy of the i386 CPU target.
     455The L4RTEMS project can be found on [https://github.com/phipse/L4RTEMS github]
     456
     457With the virtualization layer this changes.
     458RTEMS needs a library provided by L4Re to compile and then the binary can be loaded the same way.
     459So it doesn't need to be present at compile time.
     460Furthermore, it we should be able to reuse the virtualpok BSP, provided we replace libpart.a with a library provided by L4Re.
     461
    442462=  References  =
    443463